{
  "video_id": "seTDT8Rxryc",
  "channel_slug": "techwithtim",
  "channel_handle": "Tech With Tim",
  "title": "Claude Code + Nano Banana 2 = This Changes Everything",
  "duration_seconds": 995,
  "url": "https://www.youtube.com/watch?v=seTDT8Rxryc",
  "upload_date": "20260503",
  "transcript": "You've probably heard of the state of the art image\ngeneration\nmodels like Nano Banana Pro and Nano Banana 2.\nNow, these are extremely good models and you can\ncreate some really interesting images with them.\nHowever, they're a little bit inconvenient to use.\nMost of the time\nyou need to actually go to the website, type\nin the prompt and you get mediocre results.\nSo in this video, I'm going to show you how we can\ncombine nano banana specifically in an open air tube.\nYou can use any model you want with Claude Code.\nNow this gives you insane results\nbecause rather than you\nhaving to manually generate\nthe images, you can have Claude code do it for you.\nThis means that you can create stunning websites\nwith all kinds of visuals generated by nano banana.\nThis means that you can spin off\n20 generation tasks at the exact same time.\nThis means that you don't actually have to prompt\nthe image generation yourself.\nYou can have I do it for you, which gives you\na significantly better and much more detailed image.\nThis does change everything\nwhen it comes to generating images,\nand allows you to do it from one interface\nthat you're probably already\nusing, rather than going back and forth\nbetween a ton of different windows.\nNow, funny enough, when I was filming that intro,\nI was actually having Claude code\ngenerate a bunch of images for me\nand create a full website that I can link you guys to\nto make this video easier to follow,\nand you'll see that it looks something like this.\nNow what you can actually do is go to this site.\nI'll leave a link to it in the description, and it's\ngoing to have some of the resources that you'll need\nfor this video, because we're going to be downloading\nand setting up some various skills.\nNow if you have a look here,\nthese are some images that I just generated in\nlike 1 or 2 minutes using nano banana.\nAnd I didn't need to go to the website to do this.\nI literally just told Claude Code,\nas you guys can see here, I said make a site\nthat looks super sexy\nby putting four different images\nthat we generate on the site\nthat are super sleek, clean, whatever.\nAnd then it went and started\nmaking those images using nano banana,\nand then it added them directly to the site.\nAnd the way that I did\nthese is actually use something called Json prompt.\nSo you can see that it actually specified\nall of the different kind of characteristics\nor properties that it wanted for these images\nin this structured format,\nwhich gives you a significantly better image result\nthan if you just tell it like\nmake a perfume bottle, right?\nFor example,\nyou can see that we have the type, the name,\nthe description, right Mid-Ground environment,\nall of this kind of stuff.\nSo it can be super, super detailed.\nAnd this is something that you would probably\nnever write manually yourself,\nbut the Claude code can generate super quickly.\nBut anyways, I just wanted to show you a realistic\nuse case of why this is actually, well, useful\nand why you should add it to Claude Okay,\nso the first thing we're going to do here to get set\nup is we're going to go to Google AI studio,\nand we're just going to make a new API key.\nIf you don't already have an account with Google\nAI studio,\nall you're going to do is just go to this website.\nI'll leave a link to it.\nIn the description\nyou can just type Google AI studio.\nFrom there, it's\ngoing to bring you into the playground.\nFrom here, if it's your first time,\nall you're going to do is just accept the terms\nand then what we're going to do\nis go over to get API key.\nNow you're going to notice that for API key here\nwe don't have anything.\nYou're going to go to the bottom\nleft to get to this page.\nNow what we're going to do\nis just simply press on create API key.\nWe can put this in our default project\nand just name name.\nThis Gemini API key.\nOr in our case we could change it to maybe Claude\nAPI key because that's where we're going to be using.\nSo I'm going to press on Create key.\nLet's give this a second here.\nAnd it should make one for us. Okay.\nSo we can see it's made a key for us.\nWe can go ahead and copy that.\nAnd obviously don't share that with anyone else.\nNow after that\nwhat we're going to need to do is set up billing.\nThe reason we need to set a billing\nis that because we're going to be using nano\nbanana from the API,\nwe are going to need some API credit.\nSo we're going to have to at least\nhave some kind of billing so they can charge us.\nDon't worry.\nIt is pretty inexpensive to use this,\nand I believe you will get some free credits\nwhen you start using this with Google.\nBut just be aware that this is,\nyou know, a paid service.\nGenerally, and you're not able to do advanced\nimage generation completely for free.\nSo I'm going to go to set up billing.\nSo from billing setup we're just going to go through\nall of the prompts here.\nWe're going to connect a credit card and an address.\nAnd it will be good to go.\nAll right.\nSo we can see from here\nthat we've now activated the paid tier.\nGives us some information here.\nWe're going to go ahead and press on. Done.\nAnd then we're good to go.\nNow just to make sure you don't accidentally spend\ntoo much money, especially because you're going\nto be giving this API key\nto a terminal writer to an AI agent.\nWhat we're going to do is go to spend.\nWe're going to go to our projects,\nand we just change over to this,\nand we're just going to set the spending cap.\nNow for the spending cap,\nI'm just going to change this to be $10,\njust because I don't want to spend more money\nthan that.\nI highly recommend that you set this up\nto ensure that you don't accidentally go over\nwhatever it is that you do want to spend.\nThen, at any point in time, you can go\nand check your usage here and you should be able\nto see the span, the limits, the cost,\nall of that kind of stuff.\nSo from here, once we have the API key again,\njust make sure you copy it.\nAnd then we're going to move on to the next step.\nAll right. So we've got our API key here.\nAnd the next step is to download some skills and\ninstall them in Claude code so that we can use it.\nHowever before we do that I want to give a quick\nthank you to today's sponsor, who is Skyworks.\nAnd they just shipped their massive 3.0 update.\nIt's absolutely worth your attention\nif you're already spending real money\non AI like subscriptions and tokens.\nNow a lot of you guys know OpenCL,\nbut let's be real, wrestling\nwith local setups in complex\nClaude configurations is a massive headache.\nThere's Skyworks 3.0 completely removed,\nso it's the ultimate zero\nconfig OpenCL interface\nbuilt into a unified Claude workspace.\nYou get a dedicated Claude VM ready out of the box,\nwhere you don't need to set anything up,\nand you drive your entire 24 seven Claude workforce\nfrom a browser or straight from WhatsApp or telegram.\nThe second piece here is Skyworks Skills.\nNow, these aren't basic plug ins.\nThey're enterprise grade productivity skills\nthat are built right in.\nThey handle document slides, images, sheets,\nwebsites, and even multi-track video.\nBut the magic is that you don't need to use them in\nisolation.\nYou can actually stack them together.\nSo for example,\nI gave this AI one prompt plan a tech influencer\nmarketing campaign\nfor the new M5 MacBook Pro launch in Japan.\nNow it use the document scale to pull top Japanese\ntech creators with audience demographics.\nThen it use the sheet scale to build a budget matrix\nwith projected ROI and the slide scale\nto design\na fully formatted 15 page localized pitch deck.\nNow that was all in one run\nwhile I literally did nothing.\nNow next, instead of being locked into one model,\nyou can assign the best model for each task.\nSo GPT 5.4, now 5.5 Pro for heavy coding,\nClaude opus 4.6 for deep analysis, text and research,\nGemini 3.1 Pro for image and video generation\nlike we're using here.\nOr let Sky work autonomously.\nWrote the best models for each task.\nNow, if your monthly software bills have crept\npast a few hundred a month, you're replacing separate\nsubscriptions for design data analysis in website\nbuilding and easily saving 500 to $1000 per year.\nSo stop doing the heavy lifting yourself\nand let your Claude team do it for you.\nThe links in the description\nhit the link to try Sky work for free\nand plugin code ten 2026 for 20% off.\nNow let's continue with Nano Banana Pro.\nSo like I was saying,\nwhat we need to do here is add some skills to cloth.\nNow the first skill that we're going to add\nis actually the nano banana Json prompt.\nNow this is one that I just found\nfrom the MXGp market.\nYou can actually make your own,\nbut when I read through this one\nit looked fine and it was working well for me.\nBut if we read through the skill,\nyou can see that effectively.\nWhat this does is describe the client should use this\nwhenever it's going to create\nmarketing images, UI mockups, infographics, etc.\nand when not to use this skill\nfor something like this,\nand then goes through and shows\nall of the key things that we can use for generating\nthese different types of images,\nand it explains how we can translate them into Json.\nSo if we scroll down,\nyou can see that there's a bunch of different values\nthat we can use inside of the Json schema.\nLike framing, lightning,\nyou know, lighting, surface materials, etc.\nand it gives some examples of the Json schemas\nthat can be used to generate these types of images.\nNow it's not perfect.\nAnd you could definitely create a much better skill\nthan this, but it's just something to start with.\nAnd what I wanted to show you, the basic idea here is\nthat we're going to get a much better response\nwhen we use a Json schema for generating the images,\nrather than doing it ourselves.\nSo we want to have some kind of skill built\ninto Claude code that can take a normal prompt, like\nhave a girl sitting on a chair,\nyou know, drinking a beer, whatever,\nand convert it into Json so that we can be\na lot more specific with our images.\nSo this is one of the skills that will do that.\nBut again, you can create your own.\nYou can write your own Json schema.\nYou can just give an example\nand tell code to adjust it, that you want something\nto make this a little bit faster.\nNow fortunately with this particular tool\nyou can actually install these skills\njust by using this npm command.\nSo I'm just going to copy this right here.\nI'm going to go into my terminal.\nAnd of course this assumes that you have Claude code\ninstalled on your machine already.\nWhich means you'll have Node.js\nand just paste this command.\nSo it's scale fish add.\nAnd then this guy's username slash Claude code\nnano banana skills okay.\nSo we're going to go ahead and press enter.\nWhen we do that it's going to find the scale.\nIt's going to run through here.\nWe can choose where we want to install it.\nSo I'm going to go global.\nSo it's always available.\nDo you want to install it\nto all of the detected agents I'm going to go.\nYes because I have all of these on my computer.\nAnd then I'm going to go yes. And install the skill.\nNow you can see that\nI already have it installed so just skipped it.\nBut this should install it\nin all of your different agents. You'll.\nSo you just pick one that you want to add it to.\nSo now if we open up Claude. So let's go here.\nWe should be able to view our different skills.\nSo let's make this a little bit larger.\nSo we should see now that we have this Json\nprompting for nano banana.\nSo now if we reference that skill and then we ask it\nto do something it should be able to create that.\nSo actually I'll just close the terminal.\nAnd instead I'm just going to go into my Claude\ncode desktop application\nbecause I typically prefer to use this.\nAnd from here what we can do is we can just go slash\nand same thing.\nWe can use the scale.\nSo Json prompting for nano banana\nand then we can ask it to do something\njust to test that this is working.\nSo I'm just going to tell it the following.\nBased on everything you know about me, generate\nan image of me tech with Tim is sitting at a laptop,\ndrinking a Coca cola\nand working on an AI agent using Claude\nI have no idea what this is going to give me,\nbut let's see\nif will give us the result\nand then I'm just going to tell it.\nGive me the Json result.\nI want to have a look and see what you created.\nOkay. And let's press enter.\nAnd by the way, if you're wondering what I'm using to\ndictate here, I'm using a tool called Whisper Flow.\nIt's super cool.\nYou can see that I have a ton of words.\nI'm absolutely a power user here, and it gives you\nreally precise AI transcriptions extremely quickly.\nDoes automatic formatting,\npunctuation, capitalization, etc.\nand generally just allows you\nto have much better prompts\nthan if you were going to type them manually,\nbecause you can do it so much faster.\nAnyways, I'll leave a link to it in the description\nbecause I do have a long term partnership with them,\nbut I've been using it much before I worked with them\nbecause it's just such a good tool.\nAnyways,\nyou can see here that now it's giving me the Json\nand this is exactly what we were looking for,\nwhere it's giving us all of these details\nthat we want to include in the final image.\nSo this is kind of the first skill.\nI'm not going to wait for this to finish.\nThe next skill that we need to add allows us to\nactually call nano banana by using this skill first.\nSo the basic idea is we generate the Json.\nThen once the Json is done,\nwe pass that to narrow banana.\nThen it generates the images.\nAnd then from there we can stack that and combine\nthat with any other skills that we want.\nSo we can do it all directly from our terminal.\nSo now what we can do is we can go to this website\nthat I'll leave a link to in the description.\nWe can press this button for download scale.\nIt's going to give you a zip folder.\nWhat you can do is just open up the zip folder.\nAnd you can copy all of these files\nand just give it to Claude.\nSo you just paste all of these files\ninto the Claude code terminal.\nOr you can do it in, you know, however you want,\nbut essentially just give it to Claude\nand then just say something like, create a new skill\nfor generating images with nano banana.\nUsing these different files.\nDon't make any changes.\nJust wrap this in a scale\nand make sure it's available globally.\nSo all we're doing is just taking the files.\nI've already created them.\nI can verify that they work and we just give it over\nto Claude, and it should make the new skill for us.\nOkay.\nSo we can see that everything's working here now.\nI already have it installed, so just saying, hey,\nI already have all these files, but for you,\nit should create the new skill\nand then says to start using it,\nwe need to run this install command\nto install all of the dependencies.\nAnd then we need to export our Gemini API key.\nSo what I'm going to do\nis just tell it to run this command.\nSo I'm going to say run this command okay.\nAnd then the next step\nis going to be to go and get our Gemini API key.\nThen to export the API key so it's available and\nwe can actually use it to generate the images okay.\nSo we can see that\nthe install command was successful.\nNow the next thing we're going to need to\ndo is generate or export.\nSorry the Gemini API key.\nNow to get the key same thing.\nGo back to your AI studio from Google,\ngo to API keys, press on it and then simply copy it.\nNow again, don't share this with anyone else\nbecause if you do, then\nthey will be able to use the key and cost you money.\nNow if you want, you can actually just manually write\nthis command, right?\nSo export Gemini API key equals and then the key.\nHowever, if you're really lazy like me,\nyou can just paste the API key\nso you can say, you know API key like this\nokay.\nAnd then you can say something like here's\nmy API key.\nJust export it and run the command for me.\nNow, I definitely cannot recommend doing this\nbecause you're going to have your API key available\nin the chat history,\nand you're going to be giving this over to anthropic.\nYou could technically use it however they want.\nHowever, if you are super lazy, you don't care.\nThey're going to delete the key later\nthat I would just do this because it's a lot easier.\nSo effectively what you need to do is just get\nthe API key saved in your kind of system path.\nAgain, you can run this command to do it,\nor you can simply just paste it\nin the terminal and say, hey, here's my API key.\nYou know, go ahead and run this command for me.\nIt will save it.\nAnd then you'll be able to start utilizing.\nNow, in my case, my API key is already saved.\nThis one you didn't already see that\nI didn't leak in this video.\nSo I'm just going to leave it as it is.\nBut for you, you just need the API key available.\nThen you're good to go.\nAnd again, if you're having trouble\njust ask Claude code and it should be able\nto figure it out for you\nas long as you know what the actual key is.\nOkay, so now we have the Json generator.\nWe should have the Nano Banana Pro skill.\nAnd by the way,\nwhat this does is it just uses the Python API.\nI'll show\nyou right here to generate different images.\nSo you can see that there's actually documentation\nwhere you can just use Python\nand you can generate an image with, you know, Gemini\n3.1 or Gemini 3.1 Pro or whatever, right?\nYou can do multi-tenant image editing.\nYou can pass a reference image\nand then have it be updated.\nThere's a bunch of different stuff\nthat you can do here.\nAnd the script that I have in this skill\nis kind of just a minimal version of this API\nthat allows you to have a reference image\ngenerate different resolution sizes.\nYou know, what do you call aspect ratios,\nall of that kind of stuff.\nSo for now, let's start\nby asking you to generate an image.\nI would like you to generate an image for me\nusing the new skills that we created\nand added as well as the Json,\nyou know, prompting for nano banana Pro,\nI would like this image to be a nine by 16 image\nthat I could post on Instagram.\nI want it to be again a photo of me similar\nto before, where I'm kind of standing up full body\nshot, holding a sprite, and in my other hand\nI have a mac mini that has an open Claude logo on it.\nAgain, I don't know if that's actually\ngoing to make any sense, but I'm just giving it\nsome random stuff to see what it can generate.\nLet's hit enter and let's see what we get, okay.\nAnd there we go. That is the image that we get.\nNow look it's not great,\nbut at least the sprite logo is pretty good.\nI mean the person looks pretty realistic.\nI know you still kind of get a little bit of the\nAI vibe, but that is like somewhat.\nMy build obviously doesn't look exactly like me, but\nyou did get a lot of things\nright, such as the first thing right.\nThat was a really simple prompt,\nbut we can do a lot more.\nOkay, so to go a step further,\nI'm actually just going to upload this thumbnail\nthat I used for a previous video,\nwhich was also generated with AI,\nand I'm just going to ask it to change something.\nSo maybe like let's swap the Claude logo\nfor ChatGPT logo or something.\nRight?\nSo if I do that I might say something like swap\nthe Claude logo for a chat GPT logo in this image.\nAgain, use the Nano Banana Pro scale okay.\nAnd go ahead and press enter.\nAnd we should be able to use this reference image\nas well, which is something I built into the skill\nthat the way that I built this skill is,\nit actually needs a reference to the file\nin order to pass that as an input field\nfor the Python script that it's running.\nSo I just told it, it's called thumbnail\none dot png from downloads.\nI'm going to stop this.\nAnd now it's just going to execute this one\nbecause it was looking for the image.\nIt couldn't find the original source.\nSo you need to have the image downloaded.\nAnd then it should be good to go okay.\nAnd there we go.\nWe can see we get these swapped image\nI mean that's actually pretty good\nthat it was just able to do that.\nI guess it's just overlaying it on top.\nBut still I mean pretty impressive.\nAnd again you can use the reference image as well.\nAnd look there is a million\nother things you can do here. Right.\nAnd this is really just the base\nand getting it set up.\nSo now you can extend it and add it to other skills.\nBut I think with that said\nthis is a good place to leave it.\nSo this is super powerful.\nI hope you guys found some value from this.\nIf you did, make sure they like, subscribe\nto the channel and I will see you in the next one.",
  "transcript_chars": 20301,
  "ingested_at": "2026-05-21T19:03:30.737242+00:00",
  "source": "retry-no-transcript",
  "yt_meta": {
    "view_count": 22213,
    "like_count": 423,
    "channel_id": "UC4JX40jDee_tINbkjycV4Sg"
  }
}